Is there any paper or blog that discuss the effect of the dimensionality of embedding layers? The Embedding layers can be used in deep learning models, like CNN or LSTM.
You may find this blog post interesting:
He explains two reasons for using embedding layers:
1) Computational efficiency: The embedding layer helps to convert a sparse encoding to a more efficient representation.
2) Better Visualization: Relation of words could be visualized in this new multi-dimensional space